id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_ddphwyykzngl5e4d6go2eoadsu	Mark Belford	Stability of topic modeling via matrix factorization	2018	14	.pdf	application/pdf	10710	1020	68	Topic models can provide us with an insight into the underlying latent structure of a large corpus of documents. have been proposed in the literature, including probabilistic topic models and techniques based on matrix factorization. More recently, Non-negative Matrix Factorization approaches (Lee and Seung, 1999) have also been successfully applied to identify topics in unstructured text (Arora Firstly, to illustrate the issue of term instability, we consider topic models generated for r = 100 runs of randomlyinitialized NMF and LDA, with a fixed number of topics k = 7 1. Generation: Create a set of base topic models by executing r runs of NMF applied to the same corpus, represented as a document-term matrix A. NMF to this matrix to produce the final ensemble topic model. (b) Apply NMF with NNDSVD initialization to documents in A from the other (f −1) folds to generate k topics. 5. K-Fold ensemble topic modeling for matrix factorization	./cache/work_ddphwyykzngl5e4d6go2eoadsu.pdf	./txt/work_ddphwyykzngl5e4d6go2eoadsu.txt